An adaptive approach for automatic design defects detection in object-oriented systems

Abstract

Design defects in object-oriented software have a detrimental impact on quality and also increase maintenance costs. Timeous detection and identification of these flaws is, therefore, necessary to avoid adverse outcomes within the system. Metric-based heuristic frameworks have recently become popular for detecting and locating object-oriented design defects from the source code. Imperfections can occur at any level and obtaining proper threshold values is often a complicated process. To lessen this impact, an adjustable threshold methodology for metric-based design flaw detection was proposed through machine learning to compute the contribution and threshold values of each metric set. The threshold values were adjusted and adapted to conform to the software data input characteristics. Results demonstrated that the proposed method generated more appropriate threshold limits for the detection of design imperfections. Implementation was simple, less time-consuming and did not require expert knowledge compared to traditional procedures.